Example 1 : First steps with SuMMIT

Context

This platform aims at running multiple state-of-the-art classifiers on a multiview dataset in a classification context. It has been developed in order to get a baseline on common algorithms for any classification task.

Adding a new classifier (monoview and/or multiview) to the benchmark as been made as simple as possible in order for users to be able to customize the set of classifiers and test their performances in a controlled environment.

Introduction to this tutorial

This tutorial will show you how to use the platform on simulated data, for the simplest problem : vanilla multiclass classification.

The data is naively generated with a soon-to-be published multiview generator that allows to control redundancy, mutual error and complementarity among the views.

For all the tutorials, we will use the same dataset.

A generated dataset to rule them all

The dataset that will be used in the examples consists in

• 500 examples that are either

  • mis-described by all the views (labelled Mutual_error_*),
  • well-described by all the views (labelled Redundant_*),
  • well-described by the majority of the views (labelled Complementary_*),
  • randomly well- or mis-described by the views (labelled example_*).

• 8 balanced classes named 'label_1', …, 'label_8',

• 4 views named 'generated_view_1', …, 'generated_view_4',

  • each view consisting in 3 features.

It has been parametrized with the following error matrix :

	View 1	View 2	View 3	View 4
label_1	0.40	0.40	0.40	0.40
label_2	0.55	0.40	0.40	0.40
label_3	0.40	0.50	0.60	0.55
label_4	0.40	0.50	0.50	0.40
label_5	0.40	0.40	0.40	0.40
label_6	0.40	0.40	0.40	0.40
label_7	0.40	0.40	0.40	0.40

So this means that the view 1 should make at least 40% error on label 1 and 65% on label 2.

Getting started

Importing the platform’s execution function

>>> from multiview_platform.execute import execute

Understanding the config file

The config file that will be used in this example is located in multiview-machine-learning-omis/multiview_platform/examples/config_files/config_exmaple_1.yml

We will decrypt the main arguments :

• The first part regroups the basics :

  • log: True allows to print the log in the terminal,
  • name: [“summit_doc”] uses the plausible simulated dataset,
  • random_state: 42 fixes the seed of the random state for this benchmark, it is useful for reproductibility,
  • full: True means the benchmark will use the full dataset,
  • res_dir: “examples/results/example_1/” saves the results in multiview-machine-learning-omis/multiview_platform/examples/results/example_1

• Then the classification-related arguments :

  • split: 0.25 means that 80% of the dataset will be used to test the different classifiers and 20% to train them,

  • type: [“monoview”, “multiview”] allows for monoview and multiview algorithms to be used in the benchmark,

  • algos_monoview: [“decision_tree”] runs a Decision tree on each view,

  • algos_monoview: [“weighted_linear_early_fusion”, “weighted_linear_late_fusion”] runs a late and an early fusion,

  • The metrics configuration

    metrics:
      accuracy_score:{}
      f1_score:
        average:"micro"
    

    means that the benchmark will evaluate the performance of each algorithms on accuracy, and f1-score with a micro average (because of the multi-class setting).

Start the benchmark

During the whole benchmark, the log file will be printed in the terminal. To start the benchmark, run :

>>> execute()

The execution should take less than five minutes. We will first analyze the results and parse through the information the platform output.

Understanding the results

The result structure can be startling at first, but, as the platform provides a lot of information, it has to be organized.

The results are stored in multiview_platform/examples/results/example_1/. Here, you will find a directory with the name of the database used for the benchmark, here : summit_doc/ Finally, a directory with the date and time of the beginning of the experiment. Let’s say you started the benchmark on the 25th of December 1560, at 03:42 PM, the directory’s name should be started_1560_12_25-15_42/.

From here the result directory has the structure that follows :

| started_1560_12_25-15_42
| ├── decision_tree
| |   ├── generated_view_1
| |   |   ├── *-summary.txt
| |   |   ├── <other classifier dependant files>
| |   ├── generated_view_2
| |   |   ├── *-summary.txt
| |   |   ├── <other classifier dependant files>
| ├── [..
| ├── ..]
| ├── weighted_linear_late_fusion
| |   ├── <summary & classifier dependant files>
| ├── [..
| ├── ..]
| ├── train_labels.csv
| └── train_indices.csv
| ├── 1560_12_25-15_42-*-LOG.log
| ├── config_file.yml
| ├── *-accuracy_score*.png
| ├── *-accuracy_score*.html
| ├── *-accuracy_score*-class.html
| ├── *-accuracy_score*.csv
| ├── *-f1_score.png
| ├── *-f1_score.html
| ├── *-f1_score-class.html
| ├── *-f1_score.csv
| ├── *-error_analysis_2D.png
| ├── *-error_analysis_2D.html
| ├── *-error_analysis_bar.png
| ├── *-error_analysis_bar.html
| ├── feature_importances
| | ├── *-ViewNumber0-feature_importance.html
| | ├── *-ViewNumber0-feature_importance_dataframe.csv
| | ├── *-ViewNumber1-feature_importance.html
| | ├── *-ViewNumber1-feature_importance_dataframe.csv
| | ├── *-ViewNumber2-feature_importance.html
| | ├── *-ViewNumber2-feature_importance_dataframe.csv
| ├── *-bar_plot_data.csv
| ├── *-2D_plot_data.csv
| └── random_state.pickle

The structure may seem complex, but it provides a lot of information, from the most general to the most precise.

Let’s comment each file :

*-accuracy_score*.html, *-accuracy_score*.png and *-accuracy_score*.csv

These files contain the scores of each classifier for the accuracy metric, ordered with the best ones on the right and the worst ones on the left, as an interactive html page, an image or a csv matrix. The star after accuracy_score* means that it was the principal metric (the usefulness of the principal metric will be explained later). The html version is as follows :

This is a bar plot showing the score on the training set (light gray), and testing set (black) for each monoview classifier on each view and or each multiview classifier.

Here, the generated dataset is build to introduce some complementarity amongst the views. As a consequence, the two multiview algorithms, even if they are naive, have a better score than the decision trees.

The .csv file is a matrix with the score on train stored in the first row and the score on test stored in the second one. Each classifier is presented in a row. It is loadable with pandas.

A similar graph *-accuracy_score*-class.html, reports the error of each classifier on each class.

Here, for each classifier, 8 bars are plotted, one foe each class. It is clear that fore the monoview algorithms, in views 2 and 3, the third class is difficult, as showed in the error matrix.

*-error_analysis_2D.png and *-error_analysis_2D.html

In these files, one can visualize the success or failure of each classifier on each example.

Below, *-error_analysis_2D.html is displayed.

It is the representation of a matrix, where the rows are the examples, and the columns are the classifiers.

The examples labelled as Mutual_error_* are mis-classified by most of the algorithms, the redundant ones are well-classified, and the complementary ones are mixly classified.

Note

It is highly recommended to zoom in the html figure to see each row.

This figure is the html version of the classifiers errors’ visualization. It is interactive, so, by hovering it, the information on each classifier and example is printed. The classifiers are ordered as follows:

From left to right : all the monoview classifiers on the first view, all the ones on the second one, …, then at the far right, the multiview classifiers

This html image is also available in .png format, but is then not interactive, so harder to analyze.

In terms of information, this is useful to detect possible outlier examples in the dataset and failing classifers.

For example, a mainly black horizontal line for an example means that it has been missclassified by most of the classifiers. It could mean that the example is incorrectly labeled in the dataset or is very hard to classify.

Symmetrically, a mainly-black column means that a classifier spectacularly failed on the asked task.

The data used to generate those matrices is available in *-2D_plot_data.csv

*-error_analysis_bar.png and *-error_analysis_bar.html

This file is a different way to visualize the same information as the two previous ones. Indeed, it is a bar plot, with a bar for each example, counting the ratio of classifiers that failed to classify this particular example.

All the spikes are the mutual error examples, the complementary ones are the 0.33 bars and the redundant are the empty spaces.

The data used to generate this graph is available in *-bar_plot_data.csv

config_file.yml

This is a copy of the configuration file used to run the experiment.

random_state.pickle

A save of the numpy random state that was used for the experiment, it is mainly useful if no seed is specified in the config file.

1560_12_25-15_42-*-LOG.log

The log file

Classifier-dependant files

For each classifier, at least one file is generated, called *-summary.txt.

Classification on doc_summit for generated_view_1 with decision_tree.

Database configuration : 
        - Database name : doc_summit
        - View name : generated_view_1   View shape : (296, 3)
        - Learning Rate : 0.75
        - Labels used : label_1, label_2, label_3, label_4, label_5, label_6, label_7, label_8
        - Number of cross validation folds : 2

Classifier configuration : 
        - DecisionTree with max_depth : 3, criterion : gini, splitter : best, random_state : RandomState(MT19937)
        - Executed on 1 core(s) 


        For Accuracy score using {}, (higher is better) : 
                - Score on train : 0.5765765765765766
                - Score on test : 0.5540540540540541

        For F1 score using average: micro, {} (higher is better) : 
                - Score on train : 0.5765765765765766
                - Score on test : 0.5540540540540541

Test set confusion matrix : 

╒═════════╤═══════════╤═══════════╤═══════════╤═══════════╤═══════════╤═══════════╤═══════════╤═══════════╕
│         │   label_1 │   label_2 │   label_3 │   label_4 │   label_5 │   label_6 │   label_7 │   label_8 │
╞═════════╪═══════════╪═══════════╪═══════════╪═══════════╪═══════════╪═══════════╪═══════════╪═══════════╡
│ label_1 │         5 │         0 │         2 │         0 │         1 │         1 │         0 │         0 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_2 │         1 │         3 │         1 │         0 │         1 │         1 │         1 │         1 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_3 │         0 │         1 │         4 │         2 │         1 │         1 │         0 │         0 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_4 │         2 │         0 │         1 │         5 │         1 │         1 │         0 │         0 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_5 │         0 │         0 │         0 │         1 │         7 │         0 │         0 │         1 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_6 │         0 │         0 │         0 │         1 │         1 │         7 │         0 │         0 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_7 │         1 │         1 │         0 │         0 │         2 │         1 │         4 │         1 │
├─────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┼───────────┤
│ label_8 │         0 │         0 │         2 │         0 │         0 │         0 │         1 │         6 │
╘═════════╧═══════════╧═══════════╧═══════════╧═══════════╧═══════════╧═══════════╧═══════════╧═══════════╛



 Classification took 0:00:00

 Classifier Interpretation : 
First featrue : 
        2 <= 8.439512252807617
Feature importances : 
- Feature index : 0, feature importance : 0.4223852373262582
- Feature index : 1, feature importance : 0.4039223639779588
- Feature index : 2, feature importance : 0.17369239869578307

This regroups the useful information on the classifier’s configuration and it’s performance. An interpretation section is available for classifiers that present some interpretation-related information (as feature importance).